Morphological causatives are Voice over Voice Yining Nie New York University Abstract Causative morphology has been associated with either the introduction of an event of causation or the introduction of a causer argument. However, morphological causatives are mono-eventive, casting doubt on the notion that causatives fundamentally add a causing event. On the other hand, in some languages the causative morpheme is closer to the verb root than would be expected if the causative head is responsible for introducing the causer. Drawing on evidence primarily from Tagalog and Halkomelem, I argue that the syntactic configuration for morphological causatives involves Voice over Voice, and that languages differ in whether their ‘causative marker’ spells out the higher Voice, the lower Voice or both. Keywords: causative, Voice, argument structure, morpheme order, typology, Tagalog 1. Introduction Syntactic approaches to causatives generally fall into one of two camps. The first view builds on the discovery that causatives may semantically consist of multiple (sub)events (Jackendoff 1972, Dowty 1979, Parsons 1990, Levin & Rappaport Hovav 1994, a.o.). Consider the following English causative–anticausative pair. The anticausative in (1a) consists of an event of change of state, schematised in (1b). The causative in (2a) involves the same change of state plus an additional layer of semantics that conveys how that change of state is brought about (2b). (1) a. The stick broke. b. [ BECOME [ stick STATE(broken) ]] (2) a. Pat broke the stick. b. [ Pat CAUSE [ BECOME [ stick STATE(broken) ]]] Word Structure 13.1 (2020): 102–126 DOI: 10.3366/word.2020.0161 © Edinburgh University Press www.euppublishing.com/word MORPHOLOGICAL CAUSATIVES ARE VOICE OVER VOICE 103 Several linguists have proposed that the semantic CAUSE and BECOME components of the causative are encoded as independent lexical verbal heads in the syntax (Harley 1995, Cuervo 2003, Folli & Harley 2005, Pylkkänen 2008, a.o.). Each of these verbal heads (known as ‘flavours’ of v) introduces a separate event into the syntax. Therefore the anticausative in (1a) consists of a change of state event introduced by vBECOME in (3a); the causative in (2a) involves the same change of state event plus an additional causing event introduced by vCAUSE (3b). This approach to causatives, which I will refer to as the CAUSE theory, predicts that the events introduced by vCAUSE and vBECOME should be available for independent modification in the syntax. (3) a. [ vBECOME [ break stick ]] b. [ Pat vCAUSE [ vBECOME [ break stick ]]] The second view, which will be referred to as the Voice theory, takes the causative alternation to be a Voice alternation (Alexiadou et al. 2006, 2015; Schäfer 2008; Kastner 2016, 2018). In this approach, causatives are just like regular transitives in that the external argument is introduced by Voice (Kratzer 1996), sketched in (4b). The external argument is interpreted as the causer of a change of state, without introducing an additional causing event into the syntax. Thus causatives consist of a single event associated with a single v. Throughout the paper, I use the term ‘causer’ to refer to the highest argument in a causative, not to indicate a thematic role distinct from agent. (4) a. [ v [ break stick ]] b. [ Pat Voice [ v [ break stick ]]] In a language with overt causative morphology, the causative marker would spell out CAUSE in the CAUSE theory, and Voice in the Voice theory. Thus one common prediction made by these two approaches to causatives is that causative morphology should be spelled out high. That is, if affix order reflects the scopal relationships between morphemes in the syntax (Rice 2000) as expected by the Mirror Principle (Muysken 1981, Baker 1985), languages with overt voice and causative morphology should reflect the relatively high position of CAUSE or Voice. A question to be explored in this paper is whether the high causative prediction should carry over to more complex causatives. To illustrate from English, a have-causative like (5) in the CAUSE theory might be assigned a structure such as (6), where a causing event introduced by vCAUSE scopes over a doing event introduced by vDO. Under the Voice theory, which assumes that both causers and regular transitive agents are introduced by Voice, the have-causative might be assigned a structure such as (7). (5) Sam had Pat eat the cookie. (6) [ Sam vCAUSE [PatvDO [ eat [ cookie ]]]] (7) [ Sam Voice [ Pat Voice [ eat [ cookie ]]]] (6) and (7) show that for more complex causatives, the predictions of the CAUSE and Voice theories come apart. Causative morphology in the CAUSE theory should always be high. Causative morphology in the Voice theory, on the other hand, may in 104 YINING NIE principle be realised in two positions: in the higher Voice position, lower Voice position or both. This paper investigates whether the high causative prediction made by the CAUSE theory holds cross-linguistically for morphological causatives, defined as having (i) overt causative marking, (ii) a causee participant (as opposed to a theme, as in most lexical causatives), and (iii) an unmarked causative meaning (no coercion or permission reading). The high causative prediction is indeed borne out in Halkomelem, for example. As shown in the causative of an antipassive in (8), the causative affix -stǝxw, which also encodes agreement with the causee, falls outside of the antipassive affix -ǝm indicating the valency of the lower predicate. (8) niʔ cǝnq’wǝl-ǝm-stǝxw θǝ słeniʔʔǝ tθǝ sǝplil. AUX 1SBJ bake-ANTIP-CAUS.3OBJ DET woman OBL DET bread ‘I made the woman bake the bread.’ (Gerdts 2004: 769) The relative order of the causative and antipassive markers in Halkomelem reflects their syntactic and semantic scope: the causative is built on top of the antipassive. However, not all languages exhibit high causative morphology. In Tagalog, the causative marker pa- occurs inside of voice morphology, which reflects the transitivity of the entire clause. In the example in (9), the verb is prefixed with nag-, an Actor Voice (AV) marker which co-occurs with nominative case on the subject of the clause, in this case the causer. Notice that causative pa- appears closer to the verb than the nag- voice marker. Again, the relative order of the voice and causative markers reflects their syntactic scope: voice morphology is only determined once the entire causative predicate is built. (9) Nag-pa-takbo ako ng bata-ng lalaki. AV.PFV-CAUS-run 1SG.NOM GEN child-LK man ‘I made the boy run.’ (Rackowski 2002: 66) Causative morphology in Tagalog therefore appears to be low. This is unexpected in the CAUSE theory, according to which causative morphology should always be high. It is possible, however, for the low position of pa- to be captured in a Voice theory of morphological causatives. The empirical evidence points to the need for an approach to morphological causatives that allows causative morphology to appear high in some languages and low in others. I argue that the Voice theory of morphological causatives allows us to capture this cross-linguistic variation. I propose that morphological causatives involve a Voice head that selects for another VoiceP and that what has been identified as the ‘causative morpheme(s)’ in a given language can spell out either the higher Voice head, the lower one or both Voices. The causative marker in Halkomelem, for instance, spells out the higher Voice head, while the causative marker in Tagalog spells out the lower Voice head. Japanese and Kinande will be shown to spell out both. The proposed structure for causatives is given in (10), where v stands for a verbalising head v + √ROOT complex. MORPHOLOGICAL CAUSATIVES ARE VOICE OVER VOICE 105 (10) Proposed causative structure Voice2P Causer Voice2′ Voice2 Voice1P Causee Voice1′ Voice1 vP v Theme Some previous proposals have suggested that the “recursion” of Voice may be necessary in morphological causatives (Rackowski 2002, Tubino-Blanco 2010, Harley 2013). However, these approaches also assume that causative semantics arises from a syntactically-present vCAUSE head that merges above the lower VoiceP and introduces a causing event; the result is a hybrid Voice+CAUSE approach to morphological causatives, sketched in (11). (11) [ Sam Voice [ vCAUSE [ Pat Voice [ vDO [ eat [ cookie ]]]]]] My proposal takes the position that Voice over Voice is the defining property of morphological causatives, and that no dedicated CAUSE head is needed in the syntax. The present analysis assumes a Minimalist syntactic approach to word-building along the lines of Distributed Morphology (Halle & Marantz 1993), which assumes that the syntactic structure of a word is built in the syntax and gets sent to the PF and LF interfaces, where it receives a phonological form and semantic interpretation. One major conclusion of the current investigation is that morphological causatives contain no ‘causative head’ per se; rather, it is the Voice over Voice configuration that is assigned a causative interpretation at LF. Languages may choose to overtly realise different terminals in the causative configuration, which gives the impression of ‘causative morphology’ at varying heights in the structure. This kind of analysis is possible only if we assume that the determination of form and meaning follows compositional operations, as in realisational theories of morphology. The rest of the paper is structured as follows. Section 2 presents data from morphological causatives in Halkomelem (Section 2.1), whose causative marker is high, and Tagalog (Section 2.2), whose causative marker is shown to be low. Section 2.3 then provides data from languages which exhibit ‘double’ causative marking, including Japanese and Kinande. Based on this evidence, I propose in Section 3 that morphological causatives involve Voice over Voice, which captures the variable spell-out of causative marking cross-linguistically. Evidence from agenthood diagnostics supports the claim that causees in morphological causatives are indeed introduced by Voice.